Discussions about nuclear strategy have an unfortunate tendency to insulate themselves from the broader strategic context in which nuclear weapons exist.

The Interpreter's debate on the future of extended nuclear deterrence (END) has by and large avoided that pitfall. But, I wanted to look at a simple concept — proportionality  — that dictates which military tools (conventional or nuclear) will be picked out of the extended deterrence toolbox when an ally is threatened.

The concept of proportionality is central to decision-makers' calculations to use or threaten nuclear weapons. The use of nuclear weapons is a proportionate response if a security threat posed by an adversary is grave enough. If it is not, then conventional weapons or even diplomatic means might be used.

Although one of the central principles of the law of armed conflict, proportionality is not an objective calculation, as some — international lawyers in particular — believe. It depends entirely on how decision-makers perceive their security interests.

Even where nuclear weapons are reserved for existential threats, what may constitute such a threat, is hardly a settled (or short) list of possible events. The 2010 US Nuclear Posture Review used an even broader formulation, stating that the US would use its nuclear weapons to defend its 'vital interests'.

This is, of course, all common sense, but it does bring one of the major difficulties with END into sharp focus — it complicates calculations of proportionality.

Bringing a third party into a nuclear relationship means that it is no longer just two states who are forced to recognise each others' red lines. Allies often perceive threats differently and responding in a proportionate way for one ally may be viewed as an over- or under-reaction by the other.

It is not uncommon for an adversary to find and exploit this gap, where it can successfully compromise the security of one ally without provoking action by the other ally.

Differing perceptions of proportionality also explain the double-edged sword of entrapment and abandonment that a beneficiary of END typically fears — both involve a disproportionate response by the provider of END to the security threat.

Before we dump END, we need to think carefully about whether diverging post-Cold War threat perceptions among beneficiaries and the provider of END kill it or not.

Where threat perceptions align as closely as they did for the US and Europe during the Cold War, END was more effective as it offered the Soviets few gaps to exploit. Fast forward to 2010, and we see the North Koreans exploiting a gap in shelling Yeonpyeong Island.

North Korea knew the attack would harm South Korea, but that a US military response would be very unlikely. While the South Korean Defence Minister initially mooted the redeployment of tactical nuclear weapons on the Peninsula (he lost his job soon after), the US responded by sending a carrier into the Yellow Sea for military exercises. From either country's perspective, the other reacted disproportionately, just as the North Koreans would have anticipated.

If such differences emerge between provider and beneficiary that an adversary can poke holes in both the nuclear umbrella specifically and extended deterrence more broadly, how can END be anything but an abject failure, as Hyun-Wook Kim concluded? This question touches on a number of strategic dilemmas.

One of these is how we judge whether extended deterrence succeeds — is it only where the beneficiary is fully satisfied with the result, or can it succeed where the result is some sort of compromise? How does a concept such as deterrence, which requires mutual understanding, function where it involves three countries with three different perceptions of interest.

Should extended deterrence respond to 'sabre rattling' or be reserved for major threats? There are no right or wrong answers to these dilemmas, but the choice of answer determines whether one sees events like Yeonpyeong as failures, or symptomatic of the limited, imprecise but still functional nature of END.

Two brief arguments in favour of extended deterrence can be drawn out of these dilemmas, though in the Korean case they moderated rather than prevented the 'failure' (from a South Korean perspective) of extended deterrence:

First, if the alliance relationship creates an opportunity to negotiate a response that is satisfactory in both parties' opinions, bringing their calculations of proportionality closer together, then it is a plus for regional security. While the real test of this will come when the US-ROK Extended Deterrence Policy Committee has had time to work out how certain North Korean contingencies will be met, avoiding some of the more assertive options mooted in Seoul last October likely saved a fair few Korean lives.

Second, as a provider of END to multiple countries, Washington's actions are closely monitored by both allies and adversaries the world over. Maintaining its credibility to others would play no small role in its calculations of a proportionate response, giving the beneficiary of extended deterrence more clout in negotiating a response than it would otherwise have.

But two other arguments weigh in heavily against extended deterrence: First, as the Korean Peninsula situation illustrates, nuclear deterrence is usually irrelevant to crisis situations that invoke the alliance. At times, any military response at all is inappropriate, even if it would be proportionate, because of the specific risks it would create in such volatile circumstances.

Second, the likelihood of an extended deterrence failure is even higher if the provider is on an adversary's nuclear targeting list. Proportionality calculations are suddenly muddied by the desire to avoid bringing the conflict home.

Tom Mahnken and Duncan Brown seem to suggest that the US would suffer but ultimately prevail in a conflict with an adversary who could deliver nuclear weapons to US soil, an argument somewhat reminiscent of Mao's comments about China's ability to absorb a nuclear attack.

I am sceptical that a president of a democracy would take such a clinical approach to the lives of large numbers of citizens in an extended deterrence situation. Risking the blood of your citizens to protect or retaliate on behalf of your ally is an exceptionally high stakes game that is difficult to imagine today, but may well arise in the future. For example across the Taiwan Straits.

In these situations, trying to calculate a proportionate response soon reveals why END can be both a terrific success and a disastrous failure. The hard cases all depend on whether individual decision-makers are willing to let their fingers hover over the red button, and whether they think their adversary is equally willing.

Prudence dictates that we treat END as alive and well. Doing so will reduce the risk that it will ever need to be tested.

The Nuclear Reactions column is supported by the Nuclear Security Project of the Nuclear Threat Initiative, as part of a wider partnership between the NSP and the Lowy Institute.

Photo by Flickr user gravity_grave.